Agentic AI in manufacturing improves product quality and decision-making speed only when introduced with strong data foundations, layered governance, and human-in-the-loop oversight. At the AI Frontiers 2025 event, leaders from Arch Systems, ZF Group, CESMII, and Litmus outlined the specific practices that separate successful deployments from failed pilots.
“When introducing agentic AI into domain-specific workflows, what best practices help manufacturers improve product quality and decision-making speed while minimizing operational risk, especially in highly regulated or safety-critical industries?”
A panel of industry leaders—including Andrew Scheuermann, CEO of Arch Systems; Rajkumar Mylvaganan, Platform Product Manager and Head of Digital Manufacturing Platform Core at ZF Group; Jonathan Wise, Chief Technology Architect at CESMII; and Vatsal Shah, Founder & CEO of Litmus—took on this question. Their answers cut through the hype and offered practical, experience-based guidance. Here’s how they broke it down.
Building on solid foundations: the four layers
Rajkumar Mylvaganan explained that manufacturers need to think in layers. At the base are connectivity and data collection (PLCs, SCADA, MES). Above that comes data management and contextualization—using ontologies and knowledge graphs so that raw signals make sense. Only then can you add AI/analytics, and finally, agentic AI: the layer that allows systems to plan, reason, and act alongside humans.
Across all four layers runs a non-negotiable thread: governance, security, and observability. In safety-critical environments, every AI decision must be explainable, trackable, and reviewable.
Why Is Data Quality the Biggest Barrier to Agentic AI?
Poor data quality is the primary bottleneck blocking agentic AI adoption in manufacturing plants. Vatsal Shah, Founder and CEO of Litmus, stated that AI cannot “magically” clean dirty data. Manufacturers must structure existing data using common formats such as JSON, adopt standard protocols like OPC UA, and apply ontologies that describe assets consistently. Without this groundwork, agentic AI agents are set up to fail before they process a single signal.
Where Should Manufacturers Deploy Agentic AI First?
The highest-value early use cases for agentic AI are tasks where human engineers are already overstretched. Andrew Scheuermann, CEO of Arch Systems, identified three starting points: root cause analysis (RCA/CAPA), where agents scan logs and manuals to draft structured reports for operator approval; downtime classification, reducing repetitive labeling work; and unstructured data processing, turning HMI outputs and test logs into usable insights. The guiding principle: target dull, repetitive, or skipped tasks where AI improves accuracy rather than introducing new risk.
How Does Human-in-the-Loop Work for Manufacturing AI?
In regulated and safety-critical manufacturing industries, agentic AI must function as a co-pilot, not an autopilot. Andrew Scheuermann of Arch Systems described their model as giving operators a “pre-written report” generated by AI, which humans validate before any action is taken. Rajkumar Mylvaganan of ZF Group added that evaluation datasets are essential for testing agent reliability and filtering out outputs that do not meet quality thresholds. This approach keeps accountability clear while accelerating workflows.
How Does Agentic AI Work with Legacy Factory Systems?
Brownfield manufacturing environments, where machines of different ages and vendors use incompatible formats, are the norm rather than the exception. Jonathan Wise, Chief Technology Architect at CESMII, emphasized that the solution is not to rely on large language models (LLMs) to interpret inconsistent data on their own. Manufacturers must build consistency through shared semantic models and interface classes. Without standardized interpretations, AI simply moves interoperability problems up a layer rather than resolving them.
How Do Manufacturers Scale Agentic AI Beyond Pilots?
Moving agentic AI from a successful pilot to plant-wide value requires a platform approach, not a project-by-project approach. The AI Frontiers 2025 panel recommended building shared services, scalable data layers, and lifecycle management for AI agents. Change management is equally critical: operators should be involved from the start to build trust and help shape how agents learn and reason. Leadership buy-in is essential for sustaining momentum past the initial proof-of-concept stage.
What Are the Essential Building Blocks for Agentic AI?
Agentic AI delivers faster, more consistent decision-making and improved product quality in manufacturing, but only when introduced with the right foundations in place. The AI Frontiers 2025 panel identified six non-negotiable building blocks for safe, scalable deployment. These elements apply whether a manufacturer is running a first pilot or expanding across multiple plants. Each one addresses a specific failure mode that causes agentic AI initiatives to stall or produce unsafe outputs.
Essential Building Blocks:
• Standardized data and semantics
• Layered architecture with governance at every level
• Use cases targeting dull, repetitive, or skipped tasks
• Human-in-the-loop for safety and compliance
• Evaluation sets to measure and control reliability
• A platform strategy for scaling across plants
In safety-critical industries, the payoff is significant: faster, more consistent decision-making and improved product quality, delivered without sacrificing trust or control.
Related articles:
What GPT-6 Reveals About AI Readiness
Build a Factory Digital Twin in 14 Weeks
Who Owns the Risk When Factory AI Acts?
For the latest 2026 deployment data including ROI benchmarks from Siemens, Honeywell, and Rockwell Automation, see Agentic AI in Manufacturing: ROI vs. Reality.
